This study attempted to replicate and extend previous aptitude-treatment interaction (ATI) findings for students' learning in large-group and small-group teaching approaches. Two elementary teachers each taught a two-week geometry unit to two classes of fourth- and fifth-grade students. One teacher used a small-group teaching approach and the other used a large-group approach. A stratified random assignment approach was then used to ensure that within each class students were either matched or mismatched on their initial preferences for the teaching approach. Students were also stratified according to ability. At the beginning of the study students completed aptitude measures and at the end they completed achievement, attitude, and retention measures. In addition, each class was observed by a pair of observers during the course of the study. Results indicated that matching teaching approach to students' initial preferences did not affect students' achievement, retention, or attitude scores but did affect their classroom behavior. Regression analyses supported previous ATI findings in that a significant curvilinear ATI for ability was found in the present study. High and low ability students did better in the small-group approach than in the large-group approach while medium ability students did slightly better in the large-group approach than in the small-group approach. (Author/MP)